We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
PDD_GBR: Research on Evaporation Duct Height Prediction Based on Gradient Boosting Regression Algorithm.
- Authors
Zhao, Wenpeng; Li, Jincai; Zhao, Jun; Zhao, Dandan; Zhu, Xiaoyu
- Abstract
Evaporation duct is a special atmospheric stratification that always exists at sea, which has an important influence on electromagnetic wave propagation. Accurate prediction of evaporation duct height is of great significance for radio information system. In the era of big data, machine learning has been widely used in various research work and achieved a series of excellent results. Based on the observation data method, this paper studies the prediction model using Gradient Boosting Regression algorithm (GBR) and proposes the pure data‐driven GBR (PDD_GBR), GBR_Paulus‐Jeske (GBR_PJ), and GBR_Babin‐Young‐Carton (GBR_BYC) evaporation duct prediction models. Simultaneously, traditional Paulus‐Jeske (PJ) model, Babin‐Young‐Carton (BYC) model, and existing SVR_PJ model are introduced into the experiment to make a comparison. The comprehensive performance of PDD_GBR model is optimal among these models with high stability and strong generalization ability, whose prediction accuracy has a great promotion compared with other models. Key Points: We propose the pure data‐driven GBR (PDD_GBR) model of new evaporation duct prediction modelPDD_GBR model is optimal among these models in experimental area, whose prediction accuracy has a greater promotion than other modelsCompared with PJ and SVR_PJ model, RMSE is reduced by 84.6% and 54.1%, respectively, and SCC is increased by 102.3% and 40.3%, respectively
- Subjects
PARTICLE dynamics analysis; STRATIGRAPHIC geology; ELECTROMAGNETIC wave propagation; PREDICTION models; MACHINE learning
- Publication
Radio Science, 2019, Vol 54, Issue 11, p949
- ISSN
0048-6604
- Publication type
Article
- DOI
10.1029/2019RS006882